Test docker against X-ray diffraction PDBs.

[electronicsbyjulie]

Here are several PDBs of the human adenosine receptor ADORA2 bound to various ligands, made using X-ray crystallography:

https://tinyurl.com/yc7bckxx (links to rcsb.org).

Have to come up with a way to either convert these into .dock format, or better yet modify the viewer to open these PDBs natively and show the ligands in ball and stick mode. Once this is done, the crystallography models can be used to verify the accuracy of primarydock and make any necessary improvements.

[electronicsbyjulie]

ADORA2A has Phe168 in the EXR domain that in the X-ray models consistently pi-stacks with the bicyclic groups of both adenosine and caffeine. But the docker attempts to align the aromatic nitrogens with Glu13, Thr88, Asn181, Glu169, Asn253, etc. At least in the case of aromatic heterocycles, it seems pi stacking ought to have a stronger effect.

[electronicsbyjulie]

With the most recent commits of the ssepeq-acv1st branch, the pi stacking is working flawlessly however there ought to be more ionic attraction in positioning the ligand, and perhaps more tendency to favor poses with pnictogen protonation.

[electronicsbyjulie]

For the adenosine dock to work well, the functional groups of amino, aromatic nitrogenous polycycle, and polyol would have to each be considered as distinct units for the best-binding algorithm, in lieu of the current system that allows only a single heavy atom and optionally one of its hydrogens to be considered for each of the 3 positions. Then the amine could be aligned to Glu169, the heterocycles to Phe168, and the polyol to His278.

But since PrimaryDock is written mainly for olfactory reception, and adenosine is a heavier molecule than any odorant, the algorithm should be attuned to odorant ligands. Coplanar heterocycles such as pyrazines, indoles, and maltols would be capable of hydrogen bonding on their hetero atoms, while simultaneously pi stacking as conjugated rings. In these cases, the pi-stacking rings of pyrazine and indole would be considered the number 1 "most bindable" entity, and the hydrogen-bonding nitrogens/NH groups second, while the polar hydroxy group of maltol would be its first, followed by the pi ring and then the other two h-bonding oxygens. So certain atoms could be part of more than one functional group, even in cases where they may be the only atom in one of the groups.

[electronicsbyjulie]

https://www.nature.com/articles/ncomms5202

This article on hydroxyaryl aldehydes binding to and inhibiting the IRE1 protein gives results from crystallography showing the ligands pi-stacking with histidine rather than h-bonding. For olfactory purposes, vanillin would be an example of a hydroxyaryl aldehyde. Whether or not histidine would h-bond with aliphatic aldehydes, it would seem that pi-stacking an aromatic ring overrides any h-bond.

OR3A1, which is sensitive to certain aromatic aldehydes, has a couple of His residues close together in the binding pocket (His108 3.33 and His162 4.60). At first glance, these would seem the most logical place for an aldehyde group to coordinate, since they can both h-bond and pi-stack with it, albeit probably not simultaneously. There aren't many other residues in the binding pocket capable of h-bonding. Thr282 (7.42) is near Phe107 (3.32), which may function together as a combination h-bond and pi-stack C=O selector; in the AlphaFold model, Tyr181 (45.49) is also directly in the vicinity. OR3A1's agonists are aromatics with a specific length of unsaturated carbon chain, terminating in an aldehyde group, with a methyl group in the alpha or beta position. Whether the aromatic ring stacks with the His3.33 and His4.60 site or the Phe3.32 and Tyr45.49 site, the aldehyde group probably h-bonds and pi-stacks at the other site, and the length and flexibility of the aliphatic chain would be important for that. The methyl group probably creates steric effeects that favor the receptor's active configuration. It would be interesting to know whether the chirality of agonists affects OR3A1 activation.

VN1R1, by contrast, has several basic side chains in its pocket, including Lys122 (3.29), Arg129 (3.36), and Arg133 (3.40). AlphaFold places Lys3.29 as an exterior residue, and some of our alignments to equivalent OR binding site residues (Man et al) seem suspect for VN1R1, but the two arginines are definitely located in the pocket, facing TMR5, one of them forming a salt bridge with Asp228 (5.49). It's easy to imagine an aliphatic aldehyde binding strongly to this pocket.

OR10S1 is sensitive to lilial, nonanal, and at least a few of the aldehydic-smelling fatty alcohols. It has a pair of His residues at 3.33 and 4.60, a Thr at 7.42, and a Phe at 3.32, just like OR3A1. It has Phe45.49 rather than Tyr, suggesting that the hydroxyl group of tyrosine either isn't important to aldehyde binding or perhaps fulfills a ligand selectivity role. Either way, it's like looking at the same damn receptor, as if they've convergently evolved together, except in the same organism. Not many receptors have His in both the 3.33 and 4.60 positions. Phe3.32 is fairly common; Thr or Ser at 7.42 is the usual case, and an aromatic 45.49 is almost universal, so the aldehyde selectivity would seem to correlate with the double-His structure.

[electronicsbyjulie]

OR10J5 is sensitive to at least two aldehydes, lyral and bourgeonal, and to cedrene. Lyral in particular is a long molecule with an aldehyde group at one end and a hydroxy at the other. The 3D structure of OR10J5 shows Asn5.39 near His4.60, with Ser 5.43 in the same vicinity. Interestingly, the site near Thr7.42 is "wetter" than the other aldehyde receptors, with Ser7.38 and Tyr2.53 contributing extra hydroxyls to the area. This might correlate to OR10J5's sensitivity to lyral and its distal hydroxy group.

Taken together, the evidence seems to suggest that hydrophilic uncharged or basic residues capable of pi stacking, including Arg, Asn, and His, are particularly attractive to aldehydes. OR1A1's Asn109 (3.37) is thought to be important to binding its ligands, many of which are aldehydes, and the nearby Ser3.40 might contribute to the binding. But also there is evidence for aldehydes forming Michael adducts with histidine residues of proteins (here's one source: https://www.pnas.org/doi/pdf/10.1073/pnas.89.10.4544, but there are others), meaning a more complete dock would not only place the ligand into the protein, but perhaps also react it to the side chains, forming a covalent bond.